Unsupervised training in stochastically constrained STAP for nonstationary hot-clutter mitigation

摘要

This paper considers the use of "stochastically constrained" spatial and spatio-temporal adaptive processing in multimode nonstationary interference ("hot-clutter") mitigation for scenarios typical of frequency-modulated continuous waveform (FMCW) over-the-horizon radar (OTHR). In contrast to pulse waveform (PW) radar, FMCW OTHR does not usually allow access to a group of range cells that are free from the backscattered sea/terrain signal ("cold clutter"). Since supervised training methods for interference covariance matrix estimation using the cold clutter-free ranges are inappropriate in this case, we introduce and analyse adaptive routines which can operate on range cells containing a mixture of hot and cold clutter and possible targets (unsupervised training samples).